Abstract
Social network analysis (SNA) is an attractive problem for a long time when social communities were popular since 2010. Scholars like to explore the meaning behind the numerous interactions generated at these social media sites. The primary and essential issue of SNA is to monitor, estimate, and engage the potential influencers who are most relevant and active to network. If we can analyze the social network this way, business enterprises could use minimal efforts to sustain the activity of influential users, improve sales, and enhance their reputations. In this chapter, a research framework based on multiple-criteria decision making (MCDM) is proposed. The authors will show how scholars could use dynamic self-organizing map (SOM) based on multiple-objective evolving algorithm (MOEA) and static weighted influence non-linear gauge system (WINGS) to analyze a social network. Finally, comparisons are made between the innovative approaches and the methods in tradition.
TopIntroduction
Social networking sites become very popular after 2010 (Yang et al., 2016). It has been concluded that almost 98% of Internet users referring to other users’ opinions while making travel plans (Hyan Yoo & Gretzel, 2008). Therefore, if decision-makers could focus on key influencers/customers in a social network, it could take minimal effort/cost to retain their customers and promote their reputations. In the past, most related papers come from the consideration of only a single objective or criterion in Social Network Analysis (SNA). It remains a challenge or lacks the methodology to identify those active influencers in a social network by the multi-objective method. Considering the power of Multiple Criteria Decision Making (MCDM) in many academic fields, we try to find the major influencers in a social network through innovative approaches in this chapter. SNA and MCDM resolve the problem of finding key influencers and provide new insights. We review the efforts from the past and watch what is happening now. Of course,
Background
Multi-Criteria Decision Making (MCDM) methods are trendy in ranking alternatives or finding the Pareto optimum in operational research (Asghar, 2009). Nowadays, social media sites like Facebook (FB), Twitter, or Instagram are frequent fevers/fashions for business organizations to retain their loyalty. Social media sites' popularity means numerous customer data on sites are generated to form a big data set now. In the traditional ranking problem of MCDM, the number of alternatives is small. However, if we want to rank the node in a social network by MCDM; actually, the number of nodes and edges could be hundreds, thousands, even millions. In such a case, machine learning or evolutionary algorithm are valuable to solve the challenge (Graupe, 2013). The corresponding difficulty of dealing with the big data of social networks is seldom discussed in the traditional field of MCDM (Gandhi & Muruganantham, 2015).
To clarify our idea and the aim of the chapter, we present two different models via MCDM to deal with the node data from FB in the flow chart of Figure 1. First of all, the literature review of SNA and its corresponding SOM for methodology is provided. Second, the fundamentals of MCDM are reviewed, and their applications for SNA are presented. According to the literature review, it concluded that the SNA based on MCDM is relatively less. Third, we propose two different MCDM models: dynamic and static, to classify the nodes in a social network. The first one is Multi-objective SOM, a particular map from Artificial Neural Networks (ANNs) by dynamic evolution. The second one is WINGS, derived from Decision Making and Trial Evaluation Laboratory (DEMATEL), and can be widely used as a structural model for the static analysis of intertwined factors and causal relations between them. The traditional SOM is extended to minimize the weighted sum of squared errors during training; this view is original. Besides, the eigenvector centrality is famous for many years in SNA. The WINGS is innovatively used to identify the influential nodes in addition to eigenvector centrality. Finally, the computing results are compared with the traditional methodology to show the performance.
We review some necessary backgrounds in the following to help the readers know the value of the study.
Social Networks And Self Organizing Map
In this section, we review the basic knowledge for processing the SOM in social networks.
Key Terms in this Chapter
Multiobjective Programming (MOP): Multi-objective programming, also named as multi-objective optimization, vector optimization, or Pareto optimization. The MOP is a branch of operational research. The conflicting objectives form the Pareto frontier (or trade-off frontier) in the objective space. The optimal solution is considered in the presence of a trade-off frontier between two or more conflicting objectives.
Weighted Influence Non-linear Gauge System (WINGS): The WINGS method derived from DEMATEL is a branch of MCDM. WINGS is a structural model for analyzing the intertwined factors and causal relations between them. WINGS is an extension of DEMATEL, evaluates both the strength of the acting factor and the intensity of its influence; however, DEMATEL takes into consideration only the latter.
Multiple-Attribute Decision Making (MADM): MADM is a branch of MCDM. This method involves making preference decisions, such as evaluation, prioritization, and selection over the available alternatives characterized by multiple and conflicting attributes. The papers of MADM are available in many fields of ranking/selecting the appropriate alternative.
Multiple-Criteria Decision Making (MCDM): MCDM is an evaluation method to find the Pareto-optimum solution by considering many conflicting attributes or objectives at the same time. If the evaluation is based on the information from attributes, it is classified as the problem of multiple attribute decision making (MADM). If the evaluation is beased on multiobjective programming from operational research, it belongs to the multiple objective programming problems (MOPP).
Artficial Neural Network: Artificial neural networks (ANNs), also simply called neural networks (NNs), are bionic systems of neurons for vaguely computing and responding like human brains. ANNs show their power in the field of prediction and classification for a long time by a black-box system. ANNs enter a new era with the assistance of GPU for deep learning nowadays.
Self-Organizing Map (SOM): Being a particular type of ANNs, the Self Organizing Map is a simple mapping from inputs: attributes directly to outputs: clusters by the algorithm of unsupervised learning. SOM is a clustering and visualization technique in exploratory data analysis.
Multiobjective Evolutionary Algorithm (MOEA): MOEA means an evolutionary algorithm designed to resolve the MOP problem by approximating the Pareto optimum for conflicting objectives. Any algorithm that can evolve the objective value to its optimum is an evolutionary algorithm. For example, Neural Networks, Genetic Algorithms, and Particle Swarm Optimization belong to Evolutionary Algorithms (EAs) and suitable for MOEA.
Social Network: A social network is a social structure consisting of social actors (nodes) and edges between actors (nodes). The social network study aims to analyze the distinctive patterns and critical influencers observed in these structures by mathematical modeling or theories.